Re: [PATCH 1/2] vfio/type1: Adopt fast IOTLB flush interface when unmap IOVAs

[Suravee Suthikulpanit]

Hi Alex,

On 11/18/17 4:51 AM, Alex Williamson wrote:
> On Fri, 17 Nov 2017 15:11:19 -0600
> Suravee Suthikulpanit<Suravee.Suthikulpanit@xxxxxxx>  wrote:
>
> > From: Suravee Suthikulpanit<suravee.suthikulpanit@xxxxxxx>
> >
> > VFIO IOMMU type1 currently upmaps IOVA pages synchronously, which requires
> > IOTLB flushing for every unmapping. This results in large IOTLB flushing
> > overhead when handling pass-through devices with a large number of mapped
> > IOVAs (e.g. GPUs).
> Of course the type of device is really irrelevant, QEMU maps the entire
> VM address space for any assigned device.

Right, in this case certain high performance dGPUs map large address space.

> > This can be avoided by using the new IOTLB flushing interface.
> >
> > Cc: Alex Williamson<alex.williamson@xxxxxxxxxx>
> > Cc: Joerg Roedel<jroedel@xxxxxxx>
> > Signed-off-by: Suravee Suthikulpanit<suravee.suthikulpanit@xxxxxxx>
> > ---
> >   drivers/vfio/vfio_iommu_type1.c | 12 +++++++++---
> >   1 file changed, 9 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c
> > index 92155cc..28a7ab6 100644
> > --- a/drivers/vfio/vfio_iommu_type1.c
> > +++ b/drivers/vfio/vfio_iommu_type1.c
> > @@ -698,10 +698,12 @@ static long vfio_unmap_unpin(struct vfio_iommu *iommu, struct vfio_dma *dma,
> >   				break;
> >   		}
> >   
> > -		unmapped = iommu_unmap(domain->domain, iova, len);
> > +		unmapped = iommu_unmap_fast(domain->domain, iova, len);
> >   		if (WARN_ON(!unmapped))
> >   			break;
> >   
> > +		iommu_tlb_range_add(domain->domain, iova, len);
> > +
> We should only add @unmapped, not @len, right?

Hm right, I will change this in V2.

Thanks,
Suravee